Update API_Model.py

API_Model.py

@@ -21,13 +21,31 @@ model.eval().to(device)
 feature_extractor = VideoMAEFeatureExtractor.from_pretrained("MCG-NJU/videomae-base")
 label_map = {0: "Goal", 1: "Card", 2: "Substitution"}
 
-def predict_gradio(video_path):
+def predict_gradio(video):
+    import tempfile
+    import shutil
+
+    # أنشئ مجلد مؤقت للعمل
     video_id = str(uuid.uuid4())
     work_dir = f"./temp/{video_id}"
     os.makedirs(work_dir, exist_ok=True)
 
-    cap = cv2.VideoCapture(video_path)
+    # نحفظ الفيديو المرفوع على هيئة ملف مؤقت mp4
+    temp_video_path = os.path.join(work_dir, "input.mp4")
+    if isinstance(video, str):
+        # Gradio بيرسل أحيانًا مسار الملف
+        shutil.copy(video, temp_video_path)
+    else:
+        # Gradio بيرسل BytesIO stream (مش شائع بس نغطيه)
+        with open(temp_video_path, "wb") as f:
+            f.write(video.read())
+
+    # نحاول نفتح الفيديو
+    cap = cv2.VideoCapture(temp_video_path)
     fps = cap.get(cv2.CAP_PROP_FPS)
+    if fps == 0 or fps != fps:  # NaN or 0
+        return [{"error": "Invalid or unreadable video."}], ""
+
     frames = []
     while True:
         ret, frame = cap.read()
@@ -56,7 +74,7 @@ def predict_gradio(video_path):
             probs = torch.nn.functional.softmax(outputs.logits, dim=1)
             confidence, pred = torch.max(probs, dim=1)
 
-            if confidence.item() > 0.7:
+            if confidence.item() > 0.70:
                 label = label_map[pred.item()]
                 start_time = i / fps
                 end_time = min((i + segment_size), len(frames)) / fps
@@ -79,3 +97,4 @@ def predict_gradio(video_path):
         return predictions, out_path
     else:
         return predictions, ""
+